mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Adam B" <a...@mesosphere.io>
Subject Re: Review Request 36061: Slave exits gracefully on DNS lookup failure.
Date Wed, 01 Jul 2015 17:28:40 GMT


> On July 1, 2015, 12:20 a.m., Adam B wrote:
> > I'm sorry, but I don't understand how LOG(FATAL) was segfaulting before. Please
explain.
> 
> Joris Van Remoortere wrote:
>     Hi Adam,
>     This was reported as a segfault by a user in IRC. The specific segfault was fixed
by MESOS-2636.
>     The segfault _was_ happening under an error condition, so after the segfault was
fixed, we would _still_ exit with a SIGABRT / stack trace that is generated by a `LOG(FATAL)`.
>     
>     There has been a drive to reduce the number of bad user experiences by converting
some of the exit points that SIGABRT / stack trace when just a friendly message to the user
would be much more helpful.
>     I acknowledge that `PLOG(FATAL)` already prints out an error message; however, it
is rather hidden amongst the other messages and stack trace. Users tend to miss this and complain
that Mesos is broken, rather than fixing the user / configuration problem mentioned in the
error message.
>     
>     Does this answer your question?

That helps. Could somebody paste an example log/stacktrace (and exit code) from the current
behavior (without this patch), for comparison?
While I agree that we should log something more helpful than `ip.error()`, I'm not convinced
that we need to convert this (or every?) LOG(FATAL) to an EXIT().


- Adam


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/36061/#review90014
-----------------------------------------------------------


On June 30, 2015, 5:32 p.m., Marco Massenzio wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/36061/
> -----------------------------------------------------------
> 
> (Updated June 30, 2015, 5:32 p.m.)
> 
> 
> Review request for mesos, Adam B and Joris Van Remoortere.
> 
> 
> Bugs: MESOS-2962
>     https://issues.apache.org/jira/browse/MESOS-2962
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Jira: MESOS-2962
> 
> Slave fails with Abort stacktrace when DNS cannot resolve hostname
> 
> If the DNS cannot resolve the hostname for a slave node, we correctly return an Error
object, but we then fail with a segfault.
> 
> This code adds a more user-friendly message and exits normally (with an `EXIT_FAILURE`
code).
> For example, forcing `net::getIp()` to always return an Error, now causes the slave to
exit like this:
> ```
> $ ./bin/mesos-slave.sh --master=10.10.1.121:5405
> WARNING: Logging before InitGoogleLogging() is written to STDERR
> E0630 11:31:45.777465 1944417024 process.cpp:899] Could not obtain the IP address for
stratos.local; the DNS service may not be able to resolve it: >>> Marco was here!!!
> 
> $ echo $?
> 1
> ```
> 
> 
> Diffs
> -----
> 
>   3rdparty/libprocess/src/process.cpp d99947c1598c43c47c88ef3e8038081855f0d1dc 
> 
> Diff: https://reviews.apache.org/r/36061/diff/
> 
> 
> Testing
> -------
> 
> make check
> and manual failing the DNS
> 
> 
> Thanks,
> 
> Marco Massenzio
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message